Photolithography using photoresists is of considerable commercial importance in the manufacture of semiconductor devices such as integrated circuits. In typical semiconductor manufacture, a semiconductor workpiece such as an oxide covered silicon wafer is coated with a thin layer of photoresist. Selected regions of the resist-covered wafer are then exposed to light, as by passing light through a mask onto the wafer. Because the resist is photosensitive, a latent image corresponding to the exposure pattern forms in the resist layer. This latent image is developed, producing a resist mask on the underlying wafer. The mask exposes only selected areas of the wafer to chemical action, such as etching of the oxide coating, or to doping, as by ion implanatation.
There are two general types of phototresist: positive-working and negative-working. A positive resist becomes more soluble in developer upon exposure to ultraviolet light and exposed regions are washed away. A negative resist, in contrast, becomes less soluble after exposure, and the non-exposed regions are washed away.
As semiconductor technology advances to smaller and smaller devices, monitoring and fine control of the resist exposure becomes increasingly important. For positive resist, underexposure unduly broadens fine lines and, in the production of conductive regions of an integrated circuit, underexposure can limit minimum size and even produce unwanted short circuits. Similarly, overexposure can produce unwanted open circuits.
In an effort to provide better control of the exposure process, others have attempted to monitor the formation of latent images in photoresist by monitoring the light scattered by the latent image of a diffraction pattern. Preliminary results using a continuous red, helium-neon laser to monitor formation of 1.7 micron grating have been reported. As the latent image forms, it acts increasingly like a diffraction grating with a peak at the first order diffraction angle. The photoresist, which is exposed by an ultraviolet source, does not absorb red light from the laser.
This experimental approach, however, encounters technical difficulties with the smaller dimensions of leading edge technology. As linewidths approach submicron dimensions, both the spacing of the test gratings and the wavelength of the laser must become smaller and smaller. Specifically, the wavelength must be less than twice the grating spacing. But for many resists light at such short wavelengths is increasingly absorbed by the resist, producing spurious images in the photoresist and thus interfering with the parameter to be monitored.
Reducing the intensity of the monitoring radiation to levels so low that they do not expose the photoresist also reduces the signal-to-noise ratio. As a result it is extremely difficult to extract useful information from the diffracted beam. Thus a new approach is required.